Skip to content
Scintria
ToolsGrant writerFunding
HomeLiteratureClinical trialsEvidence LabProtocol packData analysisLearnProjectsLibraryReferencesApplications
Start free
ScintriaStart free

Learn / Biostatistics / Probability and distributions

The normal distribution and z-scores

13 minute read · 6-question quiz

Download this lesson as a colour Word or PDF study guide with the figures and answers. Included in paid plans; reading and quizzes stay free.Sign in

What you will learn

  • Describe the normal curve and the 68-95-99.7 rule

  • Convert a value to a z-score and use it to find probabilities and percentiles

  • Decide when normality is a reasonable assumption and what to do when it is not

Many biological measurements (height, haemoglobin, systolic blood pressure) have a bell-shaped distribution: most values are near the middle and extreme values are rare in both directions. The normal distribution is defined fully by two numbers, the mean μ and the standard deviation σ. It is symmetric about the mean, so mean, median and mode coincide.

The 68-95-99.7 rule

In any normal distribution, about 68% of values lie within 1 SD of the mean, 95% within 2 SD (more exactly 1.96), and 99.7% within 3 SD.

95% of women8.51011.51314.51617.5Haemoglobin (g/dL)
If haemoglobin in healthy adult women is normal with mean 13 and SD 1.5 g/dL, 95% lie between 10.1 and 15.9.

The z-score

A z-score says how many standard deviations a value is from the mean:

z = (x - μ)/(σ)

Converting to z puts every normal variable on a common scale (the standard normal, mean 0, SD 1), so one table or function gives all the probabilities. A z of +2 means two SDs above average; a z of −1.33 means one-and-a-third SDs below.

Worked example: anaemia in women

Haemoglobin is normally distributed with μ = 13 g/dL and σ = 1.5 g/dL. The cut-off for anaemia in non-pregnant women is 12 g/dL, but suppose we examine a more severe threshold of 11.

  1. Compute z: (11 - 13) / 1.5 = -1.33.
  2. Look up the area below z = -1.33: 0.091.
  3. So about 9.1% of women in this population would have Hb below 11 g/dL.

Now find the share between 10 and 16: z_1 = -2, z_2 = +2; the area between is about 0.954, the 95% rule again.

Finally, the 5th percentile: z = -1.645, so x = 13 - 1.645 × 1.5 = 10.53 g/dL. Five percent of women have Hb below 10.5.

9.1%-3-2-10123Hb = 11 (z = −1.33)Standard deviations from the mean (z)
A tail probability: the shaded area is the proportion of women with haemoglobin below 11 g/dL.

Three z values worth memorising

1.645 cuts off 5% in one tail (or 10% in two tails); 1.96 cuts off 2.5% in each tail (95% in the middle); 2.576 leaves 0.5% in each tail (99% in the middle). They return in confidence intervals, tests and sample size formulas.

Is my data normal?

Real data are never exactly normal, and the question is whether the approximation is good enough for the method you plan to use. Practical checks:

  • Look at a histogram and a normal quantile plot (Q-Q plot). Strong skew, a long tail or two peaks are visible at once.
  • Do not rely on a normality test alone. With a large sample, trivial departures give p < 0.05; with a small one, serious departures go undetected.
  • Know your variable. Viral load, income, length of hospital stay and many laboratory values are right-skewed. A log transform often makes them roughly normal, and the back-transformed mean is the geometric mean.

If the data cannot be made normal, use methods that do not assume it: medians with non-parametric tests, or bootstrap intervals. For means of large samples the central limit theorem (next chapter) often rescues the analysis anyway.

Reference ranges are not diagnoses

Defining "normal" as the middle 95% means that by construction 5% of healthy people are outside the range for any one test. A panel of 20 independent laboratory tests will flag at least one abnormal result in about 1 - 0.95^20 = 64\% of perfectly healthy people.

Summary

  • A normal curve is set by its mean and SD; about 95% of values lie within 1.96 SD of the mean.

  • z = (x − mean)/SD converts any normal value to the standard scale.

  • Probabilities come from areas under the curve; percentiles come from reversing the calculation.

  • Check normality visually; transform skewed data or choose a method that does not need it.

Check your understanding

Question 1
1

About what percentage of a normal distribution lies within 2 SD of the mean?

Question 2
2

Systolic BP has mean 125 and SD 15. What is the z-score of 155?

Question 3
3

Using the same distribution, about what proportion have SBP above 155?

Question 4
4

Which z value leaves 2.5% in each tail?

Question 5
5

Viral load is strongly right-skewed. A sensible first step before using a t-test is to:

Question 6
6

Why is a normality test p-value an unreliable guide on its own?

Draft study guide. Spotted an error or something unclear? Tell us at hello@scintria.com. This is educational material, not clinical or regulatory advice.

The binomial and Poisson distributionsPopulations, samples and the ways studies go wrong